Azure failed me

(this entry was previously published as a PDF on twitter since I couldn’t access my blog)
I really liked my first experience with Azure. I’m also going to certify myself (I still am). But that’s before things stopped working and I got some not very good support. This blog entry is awas a PDF because I currently can’tcan use my blog. This entry is all about why.

Today I were going to search my blog (http://blog.gauffin.org) about how to do a (almost) a generic type constraint for enums. But instead I got this screen:

image1

The scary thing is that I’ve got no notification what so ever about my site being down. So I browsed to the manage web site part of Azure and was greeted by the following screen:

image2

As you see the CPU time has been consumed. If you look at the graph you’ll see that I got a spike. Since I used a premade WordPress installation template (using the wizard in Azure) and installed it as a web site I got no control over how the CPU is used (my site is probably just a regular web site in their optimized IIS).

As I can’t see how WordPress would consume CPU in any other way than someone DOS:ing me I also started to look at the network usage for the last seven days:

image3

As you see the CPU usage is low all the time except during the spike. The strange thing is that the network usage is still normal. So it can’t really be a DOS attack or similar. Hence I also pulled the graph for the last couple of hours to make sure:

image4

Yep. No abnormal network usage.

The initial attempt to get help.

At this point there is not so much that I can do on my own. I therefore contacted my appointed support person:

Hello,

My blog spiked CPU usage so it’s now suspended.

I got three issues with that:

1. Why wasn’t I notified?
2. Why did it suddenly spike in CPU usage without any increase in network traffic? (Running wordpress)
3. I do not understand how I can increase the quota
My blogs uptime is important to me as it’s visitors steadily increases. so this down time is not welcome. Especially since it just suddenly went down.

And as a reply I got:

Jonas,

Thank you for taking the time to write to me today. As I currently see, you are currently on the 3 months free trial, here are you allowed usage: http://www.windowsazure.com/en-us/pricing/free-trial/ while on that trial. If you need to up your usage you will need to move to paying subscription

Thank you for replying. But you did not help me in any way. I know that I’ve used a trial. That should not matter. I was going to pay for it when the trial ends. I did not think that I would get suspended before that though. You could have told me “Do X to activate a paid account and your site will be up in no time”.

I tried again to say that I got a non-running site which I want up ASAP:

I’ve created a payed subscription “Användarbaserad betalning”, but the blog runs on the trial. How do I switch?

Which resulted in the following response:

You will need to open a ticket using this link: https://www.windowsazure.com/en-us/support/contact/ . For security purpose I do not have access to your account to switch that

I got no problem with that. Let’s talk with the billing department instead.

Getting help with the billing

The problem is that the technical support is not available when you are on a trial plan. But since I got no problems with starting to pay before my trial ends I contacted billing instead. I just want to get my site up and running again.

I started by emailing the exact thing as I did to my technical contact. As a response I simply got:

Hello Jonas Gauffin,

Your question appears to require technical support instead of billing and subscription support as selected

Nothing about my site being down or any attempt to help me. If they had looked at my account they should have seen that I’ve tried to convert my trial subscription to a “Pay as go” subscription. Isn’t that the minimum they could have done?

So I tried to explain to them that It’s a billing issue:

It’s billing related since I need to get it on my paid plan to get it working ASAP.

The technical issue (WHY it happened) can be solved later.

They responded with:

Hello Jonas Gauffin,
Thank you for your email. You have mentioned ‘I need to get it on my paid plan to get it working ASAP’. Could you please explain us on this and the action item for subscription management/billing team so that we could assist you further. Currently you have a ‘3-Month Free Trial’, do you mean to say you want to upgrade the trial to pay as you go.

The strange thing is that the subscription is shown as a “Pay as go” in the accounting section:

image5

(that’s Swedish for pay as go) But that’s alright, I’ll try to elaborate:

My account page says that I just got a pay to go plan.

But the hosted web is running on the trial. So how can I make it run on the payed plan?

They responded:

Hello Jonas Gauffin,

Thank you for your email. I checked your subscription 9b300662-fb0b-445a-bb89-e919908e51d2 and see that the trial period is ending on 02/28/2013. However I see that you have removed the monetary cap on this subscription, hence it is as good as a pay as you go subscription. Your subscription will automatically change to pay as you go from 02/29/2013, however if the name does not change on 29th Feb, then you could change the subscription friendly name by following the steps below:

1. Login in to https://account.windowsazure.com using your Live ID
2. Click on “Account” tab.
3. Then click on “Subscriptions” tab.
4. Click on “Edit subscription details”.
5. In the new window you can edit the name under “SUBSCRIPTION NAME”.
6. Change the subscription name and click on the tick sign to complete the process.

So what he basically is saying that I should either wait 6 hours until the next cap period starts and hope for no more CPU spikes or that I should wait one week for the new subscription period starts.

At this point I’m getting quite frustrated. I want to get my site up and running again. Hence I might have used a bit too hard tone:

I got a problem TODAY with my web site being down. Are you telling me that I should just wait one week to get it resolved?
Come on. Do you really think that that is a good answer/solution?
Is it so hard to understand that I want my website up and running again? I really do not care how or if it involves you working together with the technology department. JUST SOLVE IT.

After that I got no more replies.

Trying to migrate

After googling some I found that I can create a new site using a new paid subscription which uses the same DB and simply point my domain name to the new site.

So I created a new subscription:

image6

And then tried to create a new site using that subscription:

image7

The wizard was really easy (as last time). But once the site is being created:

image8

It’s suddenly removed and I get a notification:

image9

I tried several times and got the same error every time.

So I’ve given up doing that.

Trying to find out why I got a CPU spike

Instead I was trying to figure out why I got a CPU spike, so I started by trying to download the log file:

image10

But that didn’t work either. I got a login prompt:

image11

Hence another dead end (I did try to login with my account).

Instead I tried with my contact again:

Ok. you can still answer why the CPU usage spiked without increased network traffic.

It looks really strange when looking and the diagrams at the portal.

But no success with that either:

Jonas,

I am actually not positive as to why that happen, but you an enable logs, so then if it happens again you can see what it was

http://www.windowsazure.com/en-us/manage/services/web-sites/how-to-monitor-websites/

Yeah, right. I surely want my account to get suspended again or have to pay for unknown CPU usage.

Resignation

I give up. I simply have to wait a couple of hours so that the new quota period starts. It really sucks.
So if you see my blog up and running at http://blog.gauffin.org feel free to give me a comment and congratulate me. If not, FEEL MY PAIN!

Update (two days later)

After that I did two things:

First I posted a post in the Azure Website preview forum.

I got a great response there:

Hi, Jonas. I apologize for the trouble you’ve experienced. I looked over your document and I believe you are seeing a problem caused by a bug in WAWS. We identified it a couple of weeks ago and it’s planned to be fixed soon. (I don’t have an exact date to communicate at this point.) It’s a fairly rare thing to see, so once your quota resets, I would hope that you would not see it again.

I am concerned that you saw a failure to deploy a new site while you were running tests. If you are still seeing that problem, if you will email me at XXXXX, I can help you with that.

I appreciate your thorough documentation of this entire experience. I agree that this was not a good experience, and it’s not the experience we want our customers to have. I have downloaded the PDF you posted and I will share it with some folks here to illustrate some places where we can improve.

I also have a contact at Microsoft here in Sweden who was involved in doing a code review on a rather extensive framework which I’ve made to a client. I’ve never spoken to him since but took a long shot and contacted him. He look things up at his end and pointed me to two forum posts who discussed the same problem as I had with the CPU spikes. Lots of kudos for that.

Conclusion

If you end up having problems with Azure dont expect too much from the first line support. And you actually have to pay for technical support (even if the problem is not yours).

However, the forum which is available for all seems to be active and you get relevant answers. But then again, you can’t expect fast answers.

I wish that there had existed a proper channel to use when you are sure that the problem is not yours. They could just answer that the problem is not theirs if it’s really you that have screwed up. i.e. you only get help for things which are out of your control.

Other than that I do encourage you to try Azure out. It might have it’s child diseases, but it’s still a really smooth service to use.


  • Frank

    well what did you expect, you are running a preview service, on a trial plan. Frankly speaking it seems to be your own fault for not checking up the service agreement for a trial subscription. Although the customer support was quite shitty, it’s not all their fault.

    • JonasGauffin

      What I did expect? Well. It’s quite simple:

      1. That things **can** go wrong in previews
      2. That I get help if they do
      3. That technical support is included in trials (trials are used to evaluate the service and customer support, right?)

      I do not expect to be redirected to a new support person/forum four times. I do expect that those I contact are interested in helping me and not just fill their daily quote of answered emails.

      • Jim Cheshire

        Jonas,

        I agree with you completely. The experience you had is now what we want any of our customers to experience. I’m sorry you had such a frustrating experience.
        FYI, I told you in our forums that I would be communicating this experience to others in our organization so that we can address it. Our management team now has your original PDF documenting what happened and we will be following up.
        Thanks again.
        Jim Cheshire
        Microsoft Corporation

        • JonasGauffin

          Hello Jim,

          Thanks again for your answer. If you look at the last section of my post I’ve included the one you wrote in Azure forum.

          My last line in the conclusion also says that Azure is still a great service, although you have to work a bit with the first line support.

          I’m currently writing an article about Azure Tables (getting started article) with strategies for partition keys and row keys.

        • Dan Lineaweaver

          Jim, Microsoft needs to change their Azure support policy. It’s just terrible. All of the other providers I have used have 24/7 support built-in to their accounts, and you can get your issues resolved in one phone call. You’d think Azure would be at least as responsive. Hell, even Office 365 has Ok support – better than Azure, for sure. My experience with Azure is that, because of the weak customer service policy (there is no CS; and that horrible voice mail system – yikes!) I cannot use the service nor recommend it to any of my customers. If I can’t get my issues resolved with a phone call, I will not give you my $$$. Same with most of the customers I service. They would NEVER except these online submissions and long callback times as the only way to resolve a problem. You’re really turning-off your potential customer base with these policies, and your competition is eating your lunch. There are just so many better solutions out there! (And they’re happy to talk to you and get whomever else on the lines needed to resolve your issue.) MS, time to step up or get crushed, because right now, your business practices – at east for Azure – are pretty pathetic. Go read some books on customer service, please. I can make a few recommendations, if you’d like…

  • Craig

    I haven’t had any problems with Websites, but with Web Roles I have been able to quickly get technical people on the phone actively helping me debug problems with deployments. I found the support first rate.

  • http://twitter.com/Layershift Layershift Limited

    Sorry to hear about your troubles with Azure. I see that you’re heavily .NET focussed, so I get why you would have an interest in Azure for that kind of requirement – but curious to know if you’ve tried other PaaS that might be better suited for PHP apps like WordPress?

    We recently added PHP to our PaaS (powered by Jelastic) and would be interested to hear your perspective, as you’ve clearly got a lot of experience with Azure(!) – at least, on how our platform compares.

    (PS we definitely have real 24×7 tech. support available for all of our customers – even free trial users!)

  • http://twitter.com/EduardoMolteni Eduardo Molteni

    Being suspended without prior warning and without a clear path to normal it’s really bad and it’s one the fears that we have for moving to the cloud.

    An account suspension should be the last resort not the first.

  • Martin Horník

    Today I was surprised to find out that our company website is down without a warning! I do know that I can’t expect much from the shared plan for websites, but a warning would still be nice.
    On top of that I can’t access any details in the Azure portal and the new portal is too messy to find anything.
    A simple warning email or text message could save a lot of trouble…